- Rating
- 3(+1)
- Confidence
- 3
- Informativeness
- 3
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper proposes HalluMirror, a novel self-play framework for mitigating hallucinations in MLLLMs. An Inducer generates visual inconsistencies tailored to a Challenger's evolving weaknesses, while the Challenger learns to detect and repair them. A training-free verifier filters generated examples. Several experiments showed improved hallucination robustness. Especially, the proposed method improved performance across all eight hallucination benchmarks on GLM-4.1V-9B.
Critical strengths
- The paper addresses an important problem in mitigating hallucinations in MLLMs.
- The methodology sounds good and the methodological contribution is good for ICLR.
- A gap is addressed in adapting hallucination mitigation to the evolving weaknesses of MLLMs.
Critical weaknesses
- The performance gain is insufficient. The proposed method underperformed existing methods on the majority of benchmarks (win:loss=5:6) in Table 1.
- The proposed method should be compared with APASI[1] in Table 1.
- The hallucination verifier evaluation lacks important baselines[2][3].
- The motivation for only using existence/attribute/relation is unclear. I guess that the limited performance on FREAK and MMHal-Bench may stem from this restricted scope. Previous works [2][3][4] have considered additional hallucination types.
- [1] EMNLP 2025, https://arxiv.org/abs/2509.11287
- [2] CVPR 2025, https://arxiv.org/abs/2506.10286
- [3] CVPR 2026, https://arxiv.org/abs/2506.13130
- [4] ACL 2024, https://arxiv.org/abs/2402.03190
Questions for authors
See the critical weakness.
General Response
We sincerely thank all reviewers for their careful evaluation, constructive feedback, and thoughtful consideration. HalluMirror addresses two connected challenges in multimodal hallucination mitigation: adapting training signals to a model’s evolving hallucination weaknesses and validating automatically generated supervision against visual evidence. Below, we summarize our core contributions and the additional evidence provided during the rebuttal.
Core Contributions
Adaptive Discovery and Repair of Hallucination Weaknesses. HalluMirror couples an Inducer and a Challenger through alternating adversarial self-play. The Inducer learns to generate fine-grained hallucinations targeting the current Challenger’s failures, while the Challenger learns to detect and repair them. This interaction continually converts unresolved hallucination weaknesses into targeted training signals.
Grounding-Based Verification for Self-Play. We introduce a training-free verifier that checks existence, attribute, and relation hallucinations against visual evidence. Together with injection records and faithful reference descriptions, this supports the automatic construction of detection-and-repair supervision without hallucination-specific human annotations.
Broad Empirical Validation. Experiments cover three MLLM families and eight hallucination benchmarks. On GLM-4.1V-9B, HalluMirror improves over the base model across all eight benchmarks after three iterations, using at most 1.05K curated Challenger training examples per iteration. Evolution analyses further connect model-specific weaknesses with targeted improvements, including a 14.5% relative gain on FINER’s multi-relation subset.
Additional Evidence During the Rebuttal
We again sincerely thank all reviewers for their time, valuable suggestions, and thoughtful consideration.